PH07328: APPLICATION DELAYS OR TIMEOUTS DUE TO WRITE DRAIN LOCK HELD BY INTERNAL THREAD 010.PM2CSX01

Obtain the fix for this APAR.

APAR status

• Closed as program error.

Error description

• ERROR DESCRIPTION:
  Application delay or timeout due to write drain lock held
  by internal thread 010.PM2CSX01 which converts object to NON
  GBP-DEPENDENT.
  Dump analysis shows that PM2CSX01 triggered CASTOUT and casted
  out more pages than expected while holding drain lock. The more
  pages that need to be CASTOUT, the longer PM2CSX01 holds the
  drain lock.
  The reason that PM2CSX01 needs to CASTOUT more pages than
  expected, is because CASTOUT triggered by CLASST doesn't work
  properly. CASTOUT triggered by CLASST stopped when the number of
  changed pages was lower than about half of GBPOOLT, which left
  more than expected changed pages in the GBP for PM2CSX01 to
  CASTOUT.
  

Local fix

• LOCAL FIX:Ø
  n/a
  

Problem summary

• ****************************************************************
  * USERS AFFECTED: Db2 for z/OS V11 and V12 data sharing users. *
  ****************************************************************
  * PROBLEM DESCRIPTION: Application delays or timeouts due to   *
  *                      write drain lock held by an internal    *
  *                      thread with a correlation-ID of either  *
  *                      "010.PM2CSX01" or "014.ADB   xx".       *
  ****************************************************************
  * RECOMMENDATION:                                              *
  ****************************************************************
  Both the internal "cached state X processor" (correlation-ID
  "010.PM2CSX01") and the -ACCESS DATABASE MODE(NGBPDEP) command
  (correlation-ID "014.ADB   xx") will drive castout while holding
  a write drain when converting a pageset to non-GBP-dependent.
  For both of these operations, the intent is to avoid a long-
  running castout which would cause the write drain to affect
  other applications.
  
  It is usually expected that the normal CLASST castout will keep
  the number of changed pages in the GBP for a pageset to a
  relatively low number, or at least under the CLASST threshold.
  However, there's a problem in the castout logic that can lead
  to CLASST castout terminating much too early, leaving a very
  high number of updated pages.  If an attempt is then made to
  convert the pageset to non-GBP-dependent (before further
  castout is triggered by a threshold or a GBP checkpoint), a
  long-running castout operation may take place.
  ADDITIONAL SYMPTOMS:  MSGDSNT376I RC00C200FA on the DSOPEN lock
                        during open processing.
                        MSGDSNT501I RC00C900BA DSNILMCL
                        TYPE 00002002
  

Problem conclusion

• The castout processing for the CLASST threshold has been fixed
  to avoid terminating early.  Rather than stopping early if the
  GBPOOLT reverse threshold is hit, it will continue until the
  number of updated pages drops below 128.  This ensures that an
  operation to convert the pageset to non-GBP-dependent should
  not need to cast out more pages than the CLASST threshold.
  

Temporary fix

Comments

APAR Information

• APAR is sysrouted FROM one or more of the following:

• APAR is sysrouted TO one or more of the following:

  UI61180 UI61184

Modules/Macros

•    DSNB5PCO
  

Fix information

• Fixed component name

  DB2 OS/390 & Z/

• Fixed component ID

  5740XYR00

Applicable component levels

• RB10 PSY UI61184

     UP19/02/21 P F902

• RC10 PSY UI61180

     UP19/02/21 P F902

Fix is available

• Select the PTF appropriate for your component level. You will be required to sign in. Distribution on physical media is not available in all countries.